WS_PROJECT_SETUP.md

How to setup the Watson Studio project

The sample application requires machine learning models trained and deployed using Watson Studio. Please find step by step instruction below.

Project configuration

• create project in Watson Studio
• assosiate: Watson Machine Learning and Spark service

Training data set

• upload the training data set to DB2 Warehouse on Cloud
• in the Watson Studio create a connection to the table

Feedback data set and learning system

• upload the feedback data set to DB2 Warehouse on Cloud
• in the Watson Studio create a connection to the table

Notebook to train a model, deploy and configure payload logging

• upload a notebook to Watson Studio project

Note: Use Spark runtime when uplaoding the notebook

• use insert to code as spark df feature to insert the training data table connection (cell [2])
• replace the postgress sql database connection in payload logging section of the notebook (cell [87])
• replace wml_credentials
• run the notebook

Configure the learning system

• in the Evaluation section of the model configure learning system by providing connection to feedback table
• set the retrain option to always, redeploy if better
• run new iteration

Payload logging table and lineage

• in the studio add new connection to the payload logging table to see all scoring results logged
• the lineage can be seen on the lineage tab of the model details